Linear baseband precoding strategies for millimeter wave MIMO multi-X channels

This paper focuses on the design of linear baseband precoding and combining techniques for millimeter wave (mmWave) multiple input multiple output (MIMO) backhaul networks. We consider a scenario known as MIMO multi-X channel with L transmitter base stations and K receiver base stations, each equipped with multiple antennas, where each transmitter base station communicates simultaneously with all receiver base stations. The extremely short wavelength of mmWave requires the use of large antenna arrays at the base stations (BS) to combat the high path loss to achieve sufficiently high link gains, enabled by directional beamforming. As signal processing with large antenna arrays lead to implementations that are associated with high power consumption and high hardware complexity, hybrid (analog/digital) precoding/combining has been proposed recently to overcome these limitations. The design of such hybrid precoders/combiners is typically performed separately for the analog and the digital domains. In this work, we propose several novel low-complexity schemes suited for the digital part of such hybrid precoders/combiners. Our solutions are based on block-diagonalization and minimum mean squared error (MMSE) techniques and assume the availability of partial knowledge of CSI at each BS in the network. Finally the simulation results demonstrate the potential gains of the proposed algorithms.